A rule-based consultant for accelerator beam scheduling used in the CERN PS complex

The CERN PS accelerator complex consists of nine interacting accelerators which work together to produce particle beams for different end users, varying in particle type, energy, time structure, and geometry. The beam production schedule is time sliced, and depends on the current operational requirements, and dynamically on the accelerator status, so that production schedule changes occur in real time. Many potential schedules are not valid due to various system constraints and these constraints vary over time as new operational modes are introduced. In order to ensure that only valid schedules are given to the complex, an automated tool has been developed to indicate whether a potential schedule is valid or not. This presentation describes the method by which the validity of a beam schedule is determined and how this method was implemented using a rule-based approach based on SQL, avoiding the use of an expert system shell. Both the data to instantiate the rules and the rules themselves are kept in an Oracle data base. The SQL interpreter provides the inference engine for this knowledge based system. A few examples are presented and the running experience with the tool is discussed.